Characterization of high-explosive detonations using broadband infrared external cavity quantum cascade laser absorption spectroscopy

Measurement of temperatures and chemical species in high-explosive detonations provides critical information for optimizing explosive performance, understanding explosive effects on the surroundings, and controlling chemistry inside the explosive fireball. The process of detonation involves rapid and dramatic changes in temperature, pressure, and chemistry over multiple time and length scales, and measurement in these conditions presents numerous challenges. While a highly desirable goal is knowledge of the complete physical and chemical conditions at every point and every time during an explosive event, this is not achievable in practice and compromises must be made, often through the combination of multiple diagnostic techniques. Measurement probes such as thermocouples or piezoelectric pressure transducers are routinely used to determine time-resolved physical conditions at single points inside an explosive fireball but provide no information on chemistry. Sampling and laboratory analysis of the gaseous or solid chemical species generated after a detonation can provide detailed chemical information, but only measures the end products of the process and cannot measure transient chemical species or conditions existing during the detonation or combustion.

Optical measurement techniques are noncontact, nonintrusive, and can be performed in standoff configurations where equipment and operators are located at safe distances from the explosive event. Optical methods probing emission or absorption have an inherently fast time response usually limited only by the speed of optical detection. The explosive detonation process involves high temperatures with a corresponding strong light emission that can be analyzed to determine properties of the explosive event. High-speed imaging provides detailed information on shock wave and fireball expansion at high spatial and temporal resolution but provides limited information on chemical species.^1,2 Optical spectroscopy provides methods to measure temperature and chemical species via analysis of measured spectra, although trade-offs are required between temporal, spatial, and spectral resolution.

Optical emission diagnostics at ultraviolet (UV), visible (VIS), and near-infrared (NIR) wavelengths provide high-speed imaging and spectral information at early times of blast and fireball evolution.^3–9 Various types of point and imaging detectors are readily available for these wavelengths, which provide excellent performance in terms of speed and sensitivity. However, UV/VIS/NIR emission from explosions decreases rapidly due to fireball expansion and cooling. For explosive charges at the 10s of gram scales, optical emission at visible wavelengths typically persists for ∼10 ms, after which it is no longer detectable. Emission spectroscopy at UV/VIS wavelengths is valuable for identifying atomic species including metals and transient molecular species present at early times of shock and fireball evolution; however, these wavelengths are not ideal for measuring many important molecular species, especially for primary combustion products such as CO and CO₂.

The infrared (IR) spectral region, spanning wavelengths of ∼3–25 μm can be used to probe molecular vibrational-rotational transitions and provide numerous opportunities for chemical detection and identification. The mid-wave infrared (MWIR) region of 3–5 μm contains strong absorption features from CO, CO₂, H₂O, and NO_x species and is often used for combustion diagnostics. The long-wave infrared (LWIR) region of 8–12 μm can be used to measure a wide range of molecular species via distinctive positions and strengths of lines and bands in absorption or emission spectra and is often termed the “fingerprint” spectral region. The detection of infrared light is more challenging than at shorter wavelengths due to higher thermal background noise and limitations in detector technology, but high-speed point detectors are available. Imaging detectors in the infrared region are also available but often present limitations due to speed, array size, performance, or cost.

The infrared emission from high-explosive detonations has been measured and reported previously.^10–14 These measurements used Fourier-transform infrared (FTIR) spectroscopy to record emission in the IR spectral region of 2000–10,000 cm⁻¹. Spectrum acquisition times of 12–50 ms at spectral resolutions of 8–16 cm⁻¹ were typical for these measurements. The spectral acquisition rates are high enough to observe the expansion and cooling of the fireball for larger (10–100s kg-scale) explosions. The spectral resolution is insufficient to resolve individual lines of molecular gases, many of which have linewidths and line spacings <1 cm⁻¹, although absorption and emission bands from molecular species could be identified. The infrared emission spectrum in these studies was found to be dominated by broadband, gray-body contributions from solid particulates, with emission from bands of CO₂ and H₂O also identified in some studies.¹¹ Based on spectral modeling and fitting of the observed emission spectra, temperatures of ∼1000–2000 K were found for initial times (∼10 ms), decaying to ∼500 K within a few seconds. These studies demonstrated that valuable information was obtainable in the infrared spectral region for studying explosive fireballs.

Absorption-based optical measurements provide an alternative to emission-based measurements, allow probing of properties inside the fireball, and are also capable of measuring at times much later after detonation when optical emission has ceased. High optical density in fireball conditions usually prevents the transmission of broadband, incoherent light sources for imaging or spectroscopy, so laser-based sources are used instead. Although the use of laser-based diagnostics is routine in combustion research,^15–21 it has not seen as much application to explosive measurements. Time-resolved transmission and spectroscopy of explosives performed in the UV-VIS spectral region using continuous-wave (CW) sources or broadband modeless dye lasers have provided information on chemical species and optical opacity.^22–24 Broadband, amplified NIR light was used to measure absorption spectra and determine the temperature and water vapor concentration at a 20 kHz rate using a ruggedized probe designed to contain an open path (15 cm) of light placed within an expanding fireball.²⁵ A NIR tunable diode laser was used in a similar rugged probe to measure water vapor and temperature at a 30 kHz rate.²⁶ 

There is currently a lack of information available on MWIR and LWIR absorption properties of high-explosive fireballs, especially at early times after detonation. In addition, there is a general lack of data on the properties of detonations at later times after which emission signals are not detected. Using high-resolution absorption spectroscopy, it is possible to probe the concentrations of various molecular species present after detonation and to also determine the temperature to much lower values than possible with emission techniques. Knowledge of behavior at these late times can provide information on combustion, gas mixing, and particulate formation.

External cavity quantum cascade lasers (ECQCLs) provide a versatile infrared laser source for spectroscopy of molecules. Recent work in our laboratory has demonstrated rapid swept-wavelength tuning for broadband spectroscopy and sensing in both point and standoff measurements.^27–32 The broad tuning range (>100 cm⁻¹) enables spectroscopy of multiple species and larger molecules with broad absorption spectral features, which usually cannot be performed using traditional tunable laser spectroscopy with narrow scan ranges (<5 cm⁻¹). A high spectral scan rate (>100 Hz) combined with high spectral resolution (<0.3 cm⁻¹) enables the measurement of small molecules with narrow absorption features in transient systems and is a combination unobtainable with FTIR spectroscopy. The high spectral radiance of the ECQCL source also permits propagation over long distances or through conditions with high attenuation, such as explosive fireballs.

In this paper, we apply techniques of broadband MWIR laser spectroscopy to measure time-resolved absorption spectra during and after an explosive event. The beam from a custom-built swept-wavelength ECQCL is transmitted through an expanding explosive fireball to measure absorption along the beam path. The ECQCL wavelength is swept continuously over a tuning range of 2050–2300 cm⁻¹ (4.35–4.88 μm) to measure infrared absorption spectra at a 100 Hz rate and with a spectral resolution 0.15–0.3 cm⁻¹. High-resolution absorption spectra are acquired continuously at 10 ms intervals for times spanning a range before detonation to 100 s after detonation, and well-resolved absorption lines of CO, CO₂, H₂O, and N₂O are identified. The analysis of MWIR absorption and emission during early phases of detonation is performed with 2 μs time resolution using data acquired during spectral scans or with a fixed wavelength operation. A series of 7 explosive tests was performed using 14 g charges of PBXN-5, Comp B, PETN, and nitromethane/ammonium perchlorate (NM-AP) in an enclosed chamber. Spectral fitting of the absorption data is performed to determine fireball opacity, particulate absorption, temperature, and concentrations of CO, CO₂, H₂O, and N₂O throughout the fireball expansion and cooling. Results show a broadband attenuation from particulates throughout fireball evolution with a spectrally flat (gray-body) absorption coefficient and suggest that MWIR emission arises primarily from this particulate matter. Measured gas concentrations evolve over multiple time scales corresponding to initial expansion, cooling, and diffusive mixing. Ratios of CO:CO₂ and CO:N₂O concentrations are found to be constant over time scales 1–100 s after detonation, with variations between explosive types.

The paper is organized as follows. Section II provides details on the experimental setup for the measurements and data acquisition. Section III presents results and analysis of absorption and emission signals observed during the first 10 ms after detonation. Section IV investigates the properties of absorption and emission over longer time scales, for the fixed wavelength operation of the ECQCL. Section V presents results on spectral analysis of absorption over the duration of the explosive events. Section VI discusses noise, sensitivity, precision, and accuracy of the measurements. Section VII provides a discussion and interpretation of the measured results, and Sec. VIII is a summary.

Figures 1(a) and 1(b) show schematics of the experimental setup. A portion of the ECQCL beam was reflected from an uncoated BaF₂ wedged window (WW), transmitted through a 0.416 mm silicon etalon for wavelength scan calibration, and focused onto an infrared photodetector (MCZT 1) using a ZnSe lens (L) with 25 mm focal length. The remainder of the ECQCL beam was expanded using a lens pair with f = −50 mm and f = +75 mm, resulting in a collimated beam with diameter ∼4 mm. The collimated beam was directed through the center of an explosive blast chamber, passing through a pair of NaCl windows installed on each side of the chamber (4 windows in total). After passing through the chamber, a portion of the ECQCL beam was reflected from a BaF₂ WW and focused onto a second infrared photodetector (MCZT 2) using a ZnSe lens (L) with 25 mm focal length. The remaining ECQCL beam transmitted through the WW was directed to a beam dump (BD). The path length of the ECQCL beam inside the chamber was 1.17 m.

The explosive test chamber was designed and assembled at the Energetic Materials Diagnostic Lab at the University of Illinois Urbana-Champaign. The chamber has a volume of 2.5 m³, is sealed during detonations, and is equipped with gas purge lines to remove the gases and particulates generated during the experiment and before the chamber is reopened. Explosive charges of 14 g mass were assembled from the 4 HE types listed in Table I, selected to provide variations of combustion gas concentrations and the amount of soot production. Solid charges of PETN, PBXN-5, and CompB were pressed into cylindrical pellets. The NM-AP was a slurry of liquid nitromethane and ammonium perchlorate powder, placed into a low density polyethylene (LDPE) holder. The HE charges were detonated remotely using exploding bridgewire detonators. The solid charges were suspended inside the blast chamber at a distance of 20 cm above the ECQCL beam, as shown in the photograph in Fig. 1(c). The NM-AP slurry in its holder was placed ∼20 cm below the beam path. All detonations occurred in air at one atmosphere pressure and room temperature.

The ECQCL was assembled at the Pacific Northwest National Laboratory, with a design similar to previously reported systems.^27–32 The ECQCL utilized a QCL device designed for operation near 4.6 μm. Drive current to the QCL was modulated in amplitude from 0 to 800 mA using a 500 kHz square wave with 50% duty cycle, resulting in corresponding amplitude modulation (AM) of the output intensity from the ECQCL. Signals measured by the infrared detectors were digitized and analyzed in the software as described below.

The TE-cooled photovoltaic infrared photodetectors (VIGO PVI-4TE-6) had an active area of 0.5 × 0.5 mm² and a spectral response from ∼3 to 7 μm in wavelength. Atmospheric attenuation limits the practical detection range to the MWIR band of 3–5 μm. The detectors were coupled to transimpedance amplifiers with 200 MHz bandwidth. The signal measured by the detector contains contributions from both the ECQCL intensity transmitted through the chamber and also from MWIR light emitted by the HE fireball after detonation. Operation of the ECQCL in an AM mode provided a way to separate these two signals, similar to a method previously demonstrated using diode lasers.³³ The modulated portion of the detected signal is proportional to the ECQCL intensity, while the unmodulated portion (DC-offset) is proportional to the MWIR emission produced by the fireball. The signal measured by the MCZT detector was digitized at a 2 MHz sampling rate, thus recording 4 data points for each 500 kHz modulation cycle of the ECQCL intensity. For each modulation cycle, the peak-to-peak amplitude provides a signal proportional to the ECQCL intensity at its specified wavelength. The minimum value during each modulation cycle provides a signal proportional to the spectrally integrated fireball emission in the MWIR region. These two signals were recorded and analyzed continuously to provide the ECQCL transmitted intensity and MWIR emission intensity at intervals of 2 μs.

The ECQCL wavelength was swept repeatedly over a range of 2050–2300 cm⁻¹ (4.35–4.88 μm) using a 50 Hz sine wave applied to a galvanometer-mounted mirror inside the ECQCL cavity. Signals were recorded continuously before, during, and after each HE detonation event. Figure 2 shows an example of a 100 ms portion of data acquired before a HE detonation. The ECQCL intensity shown in Fig. 2(a) shows an overall variation in magnitude caused by the QCL gain profile as well as absorption lines from atmospheric H₂O and CO₂. The ECQCL intensity profile repeats at a 50 Hz rate due to the sinusoidal modulation of wavelength. A single 20 ms cycle shown in Fig. 2(b) contains two mirror images of the scan, corresponding to increasing and decreasing wavenumber with time. Figure 2(b) also shows the intensity transmitted through the Si etalon as a function of time. The relative wavenumber variation during the scan was calibrated using the known free spectral range of the etalon (3.488 cm⁻¹) and the absolute wavenumber was determined by comparison of measured H₂O and CO₂ lines with their expected positions obtained from the HITRAN database.³⁴ Figure 2(c) shows an example of the calibrated wavenumber vs time. Note that the edges of the scans at the extreme high and low wavenumber were discarded due to uncertainty in locating etalon fringe positions. After wavenumber calibration, the continuous ECQCL scans were subdivided into individual scans separated by 10 ms time intervals so that spectral analysis and fitting could be performed on each scan individually. The 250 cm⁻¹ (0.53 μm) scan range of the ECQCL was covered in 10 ms, giving an average scan rate of 25,000 cm⁻¹/s (53,000 nm/s).

Figure 3 shows an example of data recorded during detonation of a PBXN-5 charge. The ECQCL scan data has been calibrated in wavenumber and subdivided into scans separated by 10 ms time intervals. Before the detonation (t < 0), the scans appear similar within experimental noise and drift levels. The detonation occurs during one scan, designated by t = 0, upon which a dramatic change in intensity is observed for the ECQCL beam transmitted through the chamber. Subsequent scans (t > 0) show a reduced overall intensity due to absorption and scattering of the ECQCL intensity by solid particulates generated after detonation. In addition, as an increase in spectral absorption lines from detonation and combustion gases is observed, as will be detailed in Sec. V.

The measured ECQCL intensity was converted into absorbance units by using a portion of the data recorded before detonation for normalization: $Ai(ν~)=−ln[Ii(ν~)/I0(ν~)]$⁠. Here, $Ai(ν~)$ is the base-e absorbance of the ith scan, $Ii(ν~)$ is the measured ECQCL intensity for the ith scan, and $I0(ν~)$ is the intensity of the scan before detonation. The background $I0(ν~)$ was taken as the average of 100 scans (2 s duration) acquired in the 2 s immediately before detonation (backgrounds were calculated separately for forward and backward scans).

Figure 4 shows an example of data after conversion to absorbance. The scans before detonation are now flat lines near zero, while scans after detonation show a positive offset due to broadband particulate scattering and absorption, along with sharp spectral peaks corresponding to absorption by detonation and combustion gases. Not only does the conversion to absorbance units remove the intensity variation vs wavenumber during the scan due to the QCL gain profile but it also removes spectral features due to atmospheric constituents present along the beam path outside the chamber, which do not change during the experiment. Thus, the absorbance spectra calculated in this manner isolate the change in absorbance caused by the detonation event.

Due to the different time scales involved in the detonation event, we separate our analysis into two time periods. The first period consists of the time immediately after detonation until ∼20 ms, corresponding to the 1–2 scans during and after the detonation. For this early-time analysis, we investigate the transmission and emission variations on a fast time scale with the 2 μs point sampling time during the scans. The second time period consists of the subsequent scans after detonation (t > 10 ms). For this later time spectral analysis, we perform spectral fitting of each scan to determine the gas concentrations, temperature, and broadband attenuation due to particulates.

As noted in Sec. II, the ECQCL transmission was measured continuously throughout the detonation event; therefore, it was possible to locate the time of detonation by a corresponding dramatic change in transmitted intensity. Figure 5 shows the time dependence of the absorbance measured by the ECQCL beam and the spectrally integrated MWIR emission during the first 750 μs after detonation, for all 7 experimental runs. The t = 0 position was determined by locating a sharp spike in the signals corresponding to the electrical pickup by the detector preamplifier upon the bridgewire explosion. The ECQCL wavenumber is scanning continuously and moves by ∼20 cm⁻¹ during the time period shown, but can be considered a single wavelength source at any particular time point. It was verified that the major features in Fig. 5 do not correspond to positions of molecular spectral lines in the ECQCL scan. Because the detonation is a stochastic process, variability between tests even with the same explosive type is expected, especially at early times after breakout. Nevertheless, a number of interesting features and correlations are apparent in Fig. 5.

Because the ECQCL beam path is offset from the charge by 20 cm, there is a delay between detonation and when the effect is seen as a change in absorbance. For the traces in Figs. 5(b)–5(g), there is a sharp spike in absorbance with a duration of <10 μs observed as the first major feature after detonation. For the Comp B run in Fig. 5(a), it appears that this spike may have merged with additional absorbance features. The first spike in absorbance is most likely due to the propagation of the expanding shock wave through the ECQCL beam path, causing a transient beam deflection and corresponding reduced power on the detector. After the initial absorbance spike, additional fluctuations in absorbance are observed with varying time scales. For the case of NM-AP shown in Fig. 5(h), the absorbance rises to a constant value of ∼5 and remains high until >750 μs. For the other explosives, the absorbance fluctuates but is <4 at most times during the first 700 μs after detonation. In addition, many time periods are observed with an absorbance of ∼0 during the first 700 μs, indicating almost complete transparency. Some of the subsequent absorbance spikes appear to be correlated in time with the first absorbance spike. For example, the absorbance traces in Figs. 5(b)–5(f) show a second spike at ∼300 μs after the first absorbance spike, which could be due to a reflected shock from the chamber walls.

MWIR emission is observed immediately after detonation for all explosives except for NM-AP, which shows almost no MWIR emission. The MWIR emission is initially weak but increases gradually over the first 40 μs, resulting from light reflected and scattered from the chamber walls after detonation. For some tests [Figs. 5(a) and 5(d)–5(f)], the MWIR emission shows an initial strong peak near 100 μs after detonation with a ∼100 μs duration. Other tests [Figs. 5(b) and 5(c)] show a more gradual increase in emission intensity.

Figure 6 shows the measured absorbance and emission out to a longer time of 10 ms, plotted on a logarithmic time scale. Because the ECQCL is scanning in wavelength over this time period, the intensity may be attenuated by both broadband scattering/absorbance as well as absorption by spectral lines of molecular species, primarily CO and CO₂. Numerous molecular absorption features are visible in Fig. 6, and they appear at different times because the time axis is registered to the detonation time, which occurs at a different wavenumber scan position for each run. The molecular absorption features will be discussed in detail in Sec. V. For now, to reduce the appearance of the molecular absorption features, a smoothing filter (50 point adjacent average) was applied to the absorbance data in Fig. 6, which highlights the overall broadband absorption in this spectral region vs time, during the first 10 ms after detonation.

The MWIR emission shown in Fig. 6 over the longer time scale reveals that a second emission peak occurs in the time period of ∼0.3–10 ms for all explosives except NM-AP. In addition, there appears to be some general correlation between the emission intensity and the absorption during this time period. Comparison of the shapes of the absorption and emission vs time, especially in Figs. 6(b) and 6(c), suggests that these times correspond to high particle densities in the beam path and that the MWIR emission originates primarily from solid particles at an elevated temperature. The NM-AP shows nearly complete absorption of the ECQCL with very little MWIR emission. This observation is consistent with a cloud of cooler particles, probably resulting from the breakup of the plastic holder containing the initial liquid explosive sample. The MWIR emission has dropped to near zero by 10 ms for all tests.

To isolate the effects on absorbance solely from beam steering and particulates, a measurement of time-resolved ECQCL transmission was performed with the ECQCL not scanning, at a fixed wavenumber of 2143 cm⁻¹, corresponding to a spectral position between the P and R branches of the CO spectrum with little or no molecular absorption. Figure 7 shows the measured absorbance and emission as a function of time at a sampling rate of 2 μs for a charge of PBXN-5. The results are qualitatively similar to those shown in Figs. 5 and 6, within the limits expected for the stochastic explosive process. The emission signal shows a rise between detonation and 100 μs as more of the fireball emission is captured by the detector field of view. The emission can be detected for ∼20 ms. The absorbance shows a number of sharp spikes within the first 1 ms, most likely due to beam steering or large particles intersecting the beam path.

The measured absorbance shows large fluctuations in magnitude over multiple time scales, which decrease in both amplitude and frequency as time elapses after detonation. The initial times after detonation (t < 100 ms) are dominated by shock waves, turbulence, and other highly transient events, which would be difficult to characterize without a large number of repeated events. Later times (>100 ms) show a more uniform and consistent behavior. Figure 8(a) shows examples of the measured absorbance over 50 ms time windows spaced at various times relative to detonation. Before detonation (−1 s), the absorbance is flat with a mean of zero and a standard deviation of 0.013, corresponding to an absorbance noise of 2 × 10⁻⁵ Hz^−1/2 using a bandwidth estimated from the 2 μs time sampling. At 100 ms after detonation, the absorbance rises to a mean value of 1.5 with a standard deviation of 0.1, roughly an order of magnitude higher than the preshot fluctuations. At later times (+1 s), the absorbance has decreased to a mean value of 0.95 and the fluctuations have decreased in both amplitude and frequency, with a standard deviation of 0.05. At still later times (+10 s), the absorbance remains at a nonzero magnitude of 0.73 with a standard deviation of 0.016, slightly higher than the preshot noise.

Figures 8(b) and 8(c) show additional measurements of the mean absorbance and standard deviation over 10 ms time windows, throughout the detonation event. The absorbance rises to a peak value of ∼2 after detonation, followed by an initial rapid decay to ∼1 within 1 s. The absorbance then continues to decay at a much slower rate over a time scale of 10–100s of seconds. The fluctuations in absorbance, characterized by the standard deviation shown in Fig. 8(c), also show an initial rise followed by a fast decay before leveling off to a constant value of 0.016.

The observed behavior of absorbance vs time is consistent with an expanding volume of gases and particulates, which rapidly fill the chamber but with nonuniformities in density and possibly temperature. Based on the short-lived emission signal vs time, the average temperature of the particles decreases rapidly within the first 10 ms and emission is undetectable after 20 ms. However, the absorbance due to scattering and absorption from particulates persist for very long times after detonation due to the enclosed chamber and implying a small particle diameter that keeps the particles lofted. Based on the measurements in Fig. 8, the time to uniformly mix the particles throughout the chamber volume is ∼10 s.

As shown previously in Fig. 4, absorption lines due to molecular gases are visible within the first 10 ms after detonation and persist throughout the subsequent evolution. To analyze the absorbance spectra for the presence of various gas species, a spectral fitting procedure was performed. Absorbance spectra for CO, CO₂, and H₂O were modeled using data obtained from the HITEMP database,³⁵ and spectra for N₂O were modeled using data from HITRAN.³⁴ Standard isotopic abundances listed in HITRAN were used. To reduce computation time, lines with peak areas <0.001× the maximum peak area in the spectral region of interest were excluded from the simulation (the comparison of peak areas was performed at a temperature of 1000 K). In total, 236 lines were modeled for CO, 4938 lines were modeled for CO₂, 1066 lines were modeled for H₂O, and 725 lines were modeled for N₂O (the peak area threshold was adjusted to 0.1 for N₂O).

Because the scan resolution of the ECQCL measurement is similar to the actual spectral linewidths, there was some small but nonnegligible instrumental broadening of the measured absorbance peaks. Thus, it was not possible to extract useful information from the individual peak shapes using a Voigt fit. Instead, the individual absorption lines were modeled using Lorentzian line shapes to approximate an instrumental linewidth and also greatly reduce the computation time. All peaks of a given species were modeled using the same spectral width, and a single, uniform temperature was assumed for all species. To account for the broadband absorbance, a constant offset was added to the simulated spectra. Points in the spectrum with absorbance >5 were excluded from the fit due to low light levels on the detector. The fit was also weighted by the square of the background intensity $I0(ν~)2$ to account for the varying ECQCL power over the scan range.³² The outputs of this weighted nonlinear least squares (WNLS) fitting are the gas temperature, broadband absorbance value, and column densities NL (cm⁻²) of CO, CO₂, H₂O, and N₂O, where NL is the molecular number density N (cm⁻³) integrated over the measurement path L (cm).

Figure 9 shows example modeled absorption cross sections for CO, CO₂, H₂O, and N₂O at temperatures of 300 K and 600 K. Many absorption lines are present for each species within this spectral region, and distinct absorption bands provide unique spectral features for identification. Furthermore, the spectra have a strong dependence on temperature, indicated by the appearance of additional lines and changes in relative line areas within the bands. For the spectral fitting used here, it is the ratio of peak areas and overall band profile shape that provides information, not the individual peak shapes.

Experimental absorbance spectra were analyzed at various times after detonation. As noted previously, absorbance spectra were acquired continuously throughout the event at intervals of 10 ms. However, while this time sampling is needed at early times of the fireball evolution, it is much faster than necessary at later times when the conditions change more slowly. Therefore, a nonuniform time spacing was used in the spectral analysis corresponding to the evolving time scales of the event and to allow averaging of spectra at later times to improve the signal-to-noise ratio (SNR). For each explosive event, absorbance spectra were fit at times from 10 to 400 ms at 10 ms intervals (no averaging), from 400 to 4000 ms at 100 ms intervals (10× averaging), from 4 to 20 s at 1 s intervals (100× averaging), and from 20 to 100 s at 10 s intervals (1000× averaging).

Figure 10 shows an example of a measured absorbance spectrum obtained 1 s after detonation of Comp B, along with the best fit spectrum and fit residuals. Overall, the model spectrum reproduces the major features of the experimental spectrum. Spectral features from CO, CO₂, H₂O, and N₂O can be identified through comparison with the model spectra in Fig. 9. Features from CO (P and R branches) and CO₂ (P branch) dominate the spectrum, but N₂O (P and R branches) and H₂O (discrete lines) are also visible. The fit residuals show sharp features near the absorption lines, which results from small differences in line positions, linewidths, or line shapes between the experimental and model spectrum.²⁷ The absorption peaks measured by the ECQCL have widths ranging from 0.1 to 0.3 cm⁻¹ (full width at half maximum). The actual widths of the absorption lines fall in the range of ∼0.05–0.2 cm⁻¹, with variations between species and particular lines, confirming that the ECQCL measurement introduces an effective instrumental broadening of similar magnitude to the intrinsic peak widths.

For the spectrum in Fig. 10, obtained from Comp B at 1 s after detonation, the best fit parameters are offset = 1.05, T = 398 K, and column densities for CO = 802 × 10¹⁵ cm⁻², CO₂ = 16,260 × 10¹⁵ cm⁻², H₂O = 13,200 × 10¹⁵ cm⁻², and N₂O = 43 × 10¹⁵ cm⁻². For reference, a column density of 1 × 10¹⁵ cm⁻² corresponds to a concentration of 40 ppm for a 1 m path length of ideal gas at 298 K and 1 atm pressure. The best fit linewidths ranged from 0.17 to 0.3 cm⁻¹. The offset represents the broadband, spectrally flat absorption and scattering primarily due to soot particulates remaining after detonation. The measured gas temperature is above ambient temperature and is reasonable for a time of 1 s after detonation. The root-mean-square (RMS) sum of the fit residuals is 0.014. Errors and uncertainties in the fitting will be discussed in Sec. VI.

Figure 11 shows expanded views of various spectra to highlight features from each gas species. In all cases, the spectral model reproduces the major structure and lines in the measured spectra. Some discrepancies are apparent, primarily near the peaks of spectral lines. Figure 11(a) shows a spectral region containing the strongest H₂O features, along with several P-branch CO lines. Figure 11(b) shows a spectral region with P-branch N₂O lines interspersed with R-branch CO lines. Figure 11(c) shows the longer wavenumber region of the spectrum, which is dominated by P-branch CO₂ absorption. We note that near 300 K, most of the strongest spectral lines in this region arise from ¹³CO₂. The high total concentration of CO₂ present in both the atmosphere and after detonation results in almost complete absorption from the ¹²CO₂ lines located in the wavenumber region >2280 cm⁻¹, and thus these features are not useful for measurement. The ¹³CO₂ lines from 2230 to 2280 cm⁻¹ exhibit a good peak absorption strength for fitting due to the ∼100× lower concentration of the isotopologue. As the temperature rises, the appearance of lines from the ¹²CO₂ band begin to dominate the spectrum, as shown in Fig. 9(b).

Figure 12 shows an example of how the absorption spectrum evolves over time after detonation, for a PBXN-5 charge. At the earliest time shown of 50 ms in Fig. 12(a), there is still significant noise in the baseline due to turbulence effects, which cannot be fit using the spectral model and increases uncertainty in all fit parameters accordingly. Nevertheless, the average spectral baseline is fit by the model along with most of the spectral lines. The CO₂ lines at the long wavenumber side of the spectrum (2250–2280 cm⁻¹) are not fit as well, probably due to the high absorbance and large density of lines; however, the overall shape of the spectrum is fit. By 100 ms after detonation, shown in Fig. 12(b), the baseline fluctuations have decreased significantly, and the CO₂ lines are better fit by the model. At later times shown in Figs. 12(c) (1 s) and 12(d) (10 s), the magnitude of the spectral lines decreases, indicating a gradual reduction in species number density. The change in spectral band shapes and the number of visible absorption lines indicates a temperature decrease over time. The progression in Fig. 12 also indicates that the absorbance baseline decreases over time. At later times, the absorbance baseline is a flat line (constant magnitude with wavenumber), indicating that the particulates remaining after detonation are well represented as a gray-body, with no resonant absorption features in this wavenumber range.

Figure 13 shows the absorption spectra measured at 10 s after detonation for the 4 different explosive types. All are plotted on the same scale to highlight the differences in the gas composition. The spectral baseline is similar for Comp B, PBXN-5, and PETN, indicating a similar density of soot particulates generated from the explosives with high carbon content. The spectral baseline for NM-AP is much lower, indicating a lower particle density at this time. Significant CO is generated for all explosive types, with PBXN-5 and PETN being the highest, followed by Comp B, and with NM-AP the lowest. The CO₂ concentration is highest for PETN, followed by Comp B and PBXN-5, and is lowest for NM-AP. It is also apparent that the ratio of CO to CO₂ varies between explosive types. Another observation is that the N₂O concentration appears highest for PBXN-5, which is consistent with the higher nitrogen content of the HMX explosive used in PBXN-5.

Figure 14 shows parameters determined from fitting a series of absorption spectra in the first 400 ms after detonation. Due to the high turbulence during this time period, the fit parameters show a high variability and scatter. Furthermore, we expect differences due to the stochastic nature of the explosive process and due to the potential differences in initial conditions (e.g., exact location and orientation of explosive charge). Nevertheless, the trends and correlations in the fit parameters provide valuable information on the detonation properties and evolution with time.

The broadband absorbance (spectral baseline component) begins high and rapidly decreases, as does the gas temperature. Initial values of the broadband absorbance range from 1.5 to 3.5, decaying to 1–1.5 within 400 ms, which provides sufficient transmission for spectral measurements in the MWIR. Initial temperatures range from 700 to 1200 K, but with a high degree of uncertainty in the early times. However, within 100 ms, the temperature decreases to ∼500–550 K with much less variation between explosive types. We note that the temperature for NM-AP could not be determined reliably in this time range due to low CO and CO₂ concentrations.

The CO column density shows a variation between explosive types. Comp B, PBXN-5, and PETN show a small decay in CO column density over this time scale; however, NM-AP shows a faster decay to <5 × 10¹⁷ cm⁻² within 100 ms. The CO₂ column density was determined to be ∼2–3 × 10¹⁹ cm⁻² for all explosive types except NM-AP over this time period, with a high variability in time. The average CO₂ column density for NM-AP was an order of magnitude lower at 2 × 10¹⁸ cm⁻² over this time period. The apparent initial rise in CO₂ concentration for Comp B, PBXN-5, and PETN is not reliable due to the very high absorbance of CO₂ in the wavenumber region >2240 cm⁻¹ at the elevated temperature. It may be possible to improve the fitting for these early times by restricting the wavenumber range or weighting the spectral fit by transmittance, but these approaches have not been pursued at this time.

Figure 15 shows the variability in fit parameters for 2 runs of Comp B and 3 runs of PBXN-5. The trend in broadband absorbance is similar for all runs, within expected variations in exact particle density and mixing in the chamber. The temperatures show high variability before 100 ms, which is primarily driven by the low transmittance in the CO₂ spectral region as discussed already. However, after 100 ms, the temperatures are highly consistent between runs of the same explosive type, with a noticeable difference between Comp B and PBXN-5. For the two Comp B runs, the average temperature from 200 to 300 ms is 489 ± 9 K and 493 ± 9 K. For the three PBXN-5 runs, the average temperature from 200 to 300 ms is 446 ± 7 K, 441 ± 9 K, and 444 ± 16 K. For reference, the temperature from PETN from 200 to 300 ms is 494 ± 6 K, similar to Comp B. Thus, it appears that the gas temperature at this time window produced by PBXN-5 is lower than Comp B or PETN.

The CO column density shows a similar behavior with high reproducibility for repeated runs with the same explosive type and a difference between explosive types. The CO₂ column density was similar for all runs, although with higher variability in time. Table II summarizes the values determined for broadband absorbance, temperature, CO column density, and CO₂ column density, averaged over 200–300 ms.

Figure 16 shows the fit parameters over a longer time period from 0.1 to 100 s. The broadband absorbance from particulates continues to decrease over these time scales as the particles mix throughout the chamber and particles deposit on the chamber floor and walls. The NM-AP shows the greatest decrease, eventually reaching a value of 0.10 at 100 s. As noted above, the particulates in the case of NM-AP are primarily from the LDPE holder and are likely larger fragments that deposit more quickly than carbon soot particulates. The broadband absorbance for Comp B, PBXN-5, and PETN explosives reaches a value of ∼0.5 at 100 s, indicating that a substantial number of particulates are still present along the beam path at this time. The broadband absorbance appears to continue a decreasing trend for times beyond 100 s, which is consistent with the continued deposition of particles.

The temperatures shown in Fig. 16(b) also continue to decrease with time after the detonation. The logarithmic time scale highlights the rapid decay in temperature to ∼350 K at 5 s, followed by a much slower decay to ∼320 K at 100 s for Comp B, PBXN-5, and PETN. Over these time scales, we expect the cooling mechanism to be primarily from convection and mixing with the ambient air present in the chamber before detonation. Due to the spectral averaging used for fitting the absorbance spectra at these longer time scales, the NM-AP temperature can now be determined and is seen to be lower than the other explosives and only slightly above ambient temperatures for all times shown.

The CO and CO₂ column densities plotted in Figs. 16(c) and 16(d) show an overall decrease with time and differences between explosive types, as noted previously in Fig. 13. For each run, the CO and CO₂ column densities show a high degree of variation in time especially over longer time scales. For example, the column density for CO and CO₂ shows a local minimum for PETN near 3 s, which is not observed for other runs. These slow variations probably result from differences in how the gases mix and diffuse through the chamber, resulting in different average values along the beam path at any given time. Figure 17 plots parameters from the 3 different tests using PBXN-5 and shows fluctuations in parameters over these time scales similar to those observed between explosive types.

The time dependence of the other gases detected, H₂O and N₂O, are shown in Fig. 18. In this case, measurements of column density at t < 1 s are challenging due to low magnitude of the absorption signals and high magnitude and density of CO₂ lines. In the case of H₂O, even though the absolute number density is high, the cross sections of absorption lines in this spectral region are low. In the case of N₂O, the number density is low but can still be detected due to the high absorption cross sections of the lines. The time dependence of H₂O and N₂O shows an overall decrease in column density with time as the gases mix throughout the chamber. The H₂O column density was highest for PETN followed by Comp B and PBXN-5. No water vapor was detectable for NM-AP above measurement noise levels. The N₂O column density was highest for PBXN-5, followed by Comp B, PETN, and NM-AP. Although the typical nitrogen species of interest measured in combustion is NO, its absorbance spectrum ranges from 1700 to 2000 cm⁻¹ and could not be detected with the current ECQCL system. However, the strong absorption features of N₂O combined with the high sensitivity of the measurement allows measurement of this alternate species at lower concentrations.

Figure 19 plots the ratios of CO, N₂O, and H₂O to CO₂ column density from 1 to 100 s for the different explosive types, including the duplicate runs of the same type. Ratios for NM-AP had a high degree of scatter and uncertainty due to the low concentrations of CO₂ and other gases and are off the scale shown in the figure. Despite the large variation in absolute column densities with time, the ratio is nearly constant for CO:CO₂ and N₂O:CO₂ over the time period 1–100 s. This observation provides further evidence that the slow variations in CO, N₂O, and CO₂ column densities result from mixing in the chamber and not a chemical process. The H₂O:CO₂ ratio shows a continued decrease over time, indicating that the H₂O vapor may condense on the chamber walls or be adsorbed onto particles.

The ratio of CO:CO₂ is significantly different between the explosive types. The CO:CO₂ ratio is highest for PBXN-5, followed by PETN, and is lowest for Comp B. The duplicate runs of Comp B produce almost identical CO:CO₂ ratios, but the duplicate runs of PBXN-5 show some variability. The N₂O:CO₂ ratio is also different between explosive types, is highest for PBXN-5, followed by Comp B, and is lowest for PETN. However, the variability in N₂O:CO₂ is somewhat high for duplicate runs. Table III summarizes the mean values and standard deviations of various parameters over the time period from 10 to 100 s. The measurement path length of 117 cm was used to convert broadband absorbance to an attenuation coefficient.

Gas concentrations reported in FTIR emission studies of TNT explosions were in the range of 10¹⁷–10¹⁸ cm⁻³ for CO₂ and H₂O and 10¹⁵–10¹⁷ cm⁻³ for CO,¹³ and the results here are of similar orders of magnitude. For comparison with atmospheric concentrations of the molecular species, the detected concentrations of explosive products for the high carbon explosives are CO of ∼200 ppm, CO₂ of ∼3000 ppm, H₂O of ∼1000 ppm, and N₂O of ∼10 ppm. The measured particle attenuation coefficients are slightly higher than previously reported values of ∼10⁻²–10⁻¹ m⁻¹ obtained from FTIR emission spectroscopy of TNT explosions,¹³ but this may be related to the use of a confined chamber in the studies here.

There are multiple potential sources of uncertainty and error in measurement and analysis. The goal in the work presented here was to demonstrate a new technique for infrared laser spectroscopy in explosive detonations, and it did not include a comprehensive analysis of all error sources. Nevertheless, in this section, we present an initial characterization of measurement noise, measurement precision, and accuracy errors introduced by the spectral fitting algorithm.

The measurement noise is characterized based on the minimum detectable change in absorbance, which depends on the measurement bandwidth. A convenient way to express absorbance noise as a function of averaging time is via an Allan-Werle deviation analysis, showing the point-to-point variance (or deviation) of a series of data with progressively larger averaging bins.³⁶ Figure 20 shows the results of an Allan deviation analysis performed on the absorbance measured for fixed wavelength operations where data were acquired at 2 μs sampling intervals. A series of 5 × 10⁶ sequential points was used from a 10 s time period before detonation. The results show the minimum detectable absorbance change as a function of averaging time. The Allan deviation is 0.012 at the fastest sampling time of 2 μs, decreasing to a minimum of 3 × 10⁻⁴ for averaging times 0.01–1 s. These values are consistent with typical laser-based absorption measurements.³⁷ The peak in the Allan deviation near 0.1 ms is most likely caused by residual motion of the galvanometer scanners used in the ECQCL, which have a small angle settling time of ∼0.2 ms.

For analysis of the noise when measuring absorbance spectra, predetonation data were used again. A series of spectral scans acquired at 100 Hz over a 10 s time period before detonation was used to calculate a series of absorbance spectra, which in the absence of noise and drift would ideally be flat lines with zero mean. Figure 21(a) shows examples of predetonation absorbance spectra for various averaging factors (1, 10, 100, and 1000), corresponding to total acquisition times per averaged spectrum of 10 ms, 100 ms, 1 s, and 10 s, respectively. The absorbance noise was defined as the standard deviation of the spectrum, and as expected, the noise decreases as sequential spectra are averaged. Figure 21(b) shows the absorbance noise as a function of averaging time. For the fastest spectral acquisition time of 10 ms, the absorbance noise is 0.013. This value is nearly equal to the point-to-point Allan deviation, indicating that the scanning of ECQCL wavelength does not introduce substantial excess noise. Averaging of spectra continues to decrease the absorbance noise to a value of 1.1 × 10⁻³ for a 10 s averaging time. Overall, the results indicate excellent performance when averaging absorbance spectra to reduce noise, which allows scaling of the analysis to match the varying time scales of the experiment.

Characterization of the absorbance spectral noise is useful for evaluating system performance and can help indicate the minimum spectral absorbance features that can be detected. However, to determine the precision of the measurement, it is important to also consider the effects of the spectral analysis algorithms. To evaluate measurement precision, we performed spectral analysis on a series of spectra obtained at a time 100 s after detonation of a PBXN-5 charge when conditions are changing slowly, and for various averaging factors of 1–1000×. For each averaging factor, a set of 20 spectra (after averaging) was analyzed and the standard deviation of the obtained fit parameters was then used to determine the measurement precision.

Table IV summarizes the results on measurement precision. In general, we expect precision to improve with averaging if random noise dominates the signal. Here, the values should be considered upper limits because the actual parameters being measured (e.g., gas concentrations) may change over the measurement duration, especially for the longer averaging times. This behavior is most noticeable for the broadband absorbance, where the standard deviation appears to increase with the averaging time; however, it was verified that this was due to a steady decrease in the value of the spectral baseline parameter over the time scales investigated. The temperature shows very good precision, dropping below 1 K for a 1 s averaging time. The column density precisions improve up to a 1 s averaging time, but most increase for 10 s due to drifts in the measured parameters. The ratios of column density CO:CO₂ and N₂O:CO₂ reduce the effects of drifts in the parameter values and show a better improvement in precision with increased averaging. It is also useful to consider relative precision for typical values of the measured quantities. For a CO:CO₂ ratio of 50 × 10⁻³ and an averaging time of 1 s, the relative precision is 0.6%. For an N₂O:CO₂ ratio of 5 × 10⁻³ and 1 s averaging, the relative precision is 0.4%. The high precision of the measurements reflects the high scan reproducibility of the ECQCL system, which allows effective averaging of sequential scans to reduce noise.

The measurement accuracy is more difficult to evaluate without an external calibration by measurement of sources with known concentrations under similar and controlled physical conditions. Prior measurements with similar ECQCL systems have demonstrated excellent agreement with quantitative library spectra acquired using FTIR systems that have been rigorously calibrated.³⁸ Analysis of path-integrated spectra has inherent uncertainties due to variations of physical parameters along the path, which will be highest at early times of fireball evolution and decrease over time as conditions become more uniform. Other sources of measurement error may arise from calibration of the wavenumber axis, and nonuniform spectral broadening introduced by the ECQCL. The low noise and high precision of the measurements and improvement in averaging indicate that the ECQCL scans are highly reproducible; however, there may be constant and systematic errors present that affect the measurement accuracy.

Errors introduced by the fitting algorithm were evaluated by analyzing a series of absorption spectra simulated using HITEMP and HITRAN parameters under conditions similar to those observed in the measurements. The spectral line shapes were simulated as Voigt profiles with calculated Doppler and pressure broadening from the database parameters. Spectra were simulated from 300 to 1000 K in 100 K increments with fixed column densities and a broadband spectral offset. The simulated spectra were then analyzed using the same WNLS algorithm used to analyze the experimental data, and the fit parameters determined were compared against the parameters used to simulate the spectra.

Figure 22 shows an example of applying the WNLS fitting algorithm to synthetic spectra simulated at T = 400 K and T = 1000 K, and with other input parameters listed in the caption. The apparent scatter in the synthetic spectrum results from the discrete wavenumber sampling grid used with 0.05 cm⁻¹ spacing, corresponding to the grid used for the experimental data. The fits to the synthetic data appear qualitatively similar to the fits performed on the ECQCL experimental data. In particular, the residual structure for CO near each peak shows the effect of mismatch between the Lorentzian peaks with fixed width in the WNLS algorithm and the Voigt profiles with varying widths across the band in the synthetic spectra. The observation that the fit residuals appear similar for fitting of both synthetic and experimental data suggests that the ECQCL accurately measures the absorption spectrum, within the limitations of scan resolution and the spectral sampling grid. It also suggests that the WNLS fit accuracy could be improved by using more physically accurate Voigt profiles convolved with an empirically determined instrument function for the ECQCL scan. However, the increase in complexity of the fitting functions would need to be balanced against longer computation times for spectral fits.

The fit parameters determined from the WNLS fits were compared against the parameters used to simulate the spectra. The comparison indicates that the errors in determining the spectral offset were negligible. The errors in temperature were <10% over the 300–1000 K range. Fit errors in determining CO column density were <10% up to 500 K and then steadily increased up to 60% at 1000 K. Fit errors in CO₂ column density were <2% up to 600 K and <12% from 600 to 1000 K. Fit errors in H₂O column density were <20% over the 300–1000 K range. Fit errors in N₂O column density were <10% up to 400 K and then increased dramatically up to 1000% at T = 1000 K. The poor performance of determining N₂O at higher temperatures results from the small absorbance peaks being overwhelmed by much stronger overlapping CO₂ peaks at higher temperatures.

Overall, the WNLS fitting algorithm performs adequately from temperatures 300 to 500 K, corresponding to times >100 ms after detonation, with errors <10% in CO column density and <2% in CO₂ column density. The weaker absorption features from H₂O and N₂O are more sensitive to fitting errors, especially at higher temperatures, but errors were <20% over temperatures 300–400 K. It is apparent that significant errors are introduced by the WNLS fitting algorithm at higher temperatures, which is not surprising considering the increasing Doppler widths and decreasing Lorentzian widths of the peaks as the temperature increases will cause increasing deviations from the Lorentzian line profile used in the WNLS algorithm. In addition, the spectral congestion at higher temperatures due to strong and overlapping CO₂ absorption lines may decrease the ability of the algorithm to find an optimal fit.

Because the ECQCL beam path is located 20 cm from the initial charge and because the measurements are path-integrated along the beam path, interpretation of the absorbance and emission data at early times after detonation benefits greatly from additional knowledge of the fireball expansion. Figure 23 shows data obtained using visible high-speed imaging to measure the initial fireball expansion. The measured fireball radius is plotted vs time up to 40 μs. To estimate the dynamics of the shock wave and fireball radius, we follow a procedure detailed by Gordon et al.¹ The solid line shows a fit to the data using a drag model $R(t)=Rm[1−e−kt]$⁠, where $Rm$ is the radius of the stagnated fireball and k is the drag coefficient. Based on the drag model fit with R_m = 23 cm and k = 0.03 μs⁻¹, the fireball reaches the ECQCL beam position in ∼76 μs. The dashed line shows a fit to the data using a blast model $R(t)=atb$ where a is related to the energy released and $b=(s+2)/(n+2)$⁠; n is the expansion dimensionality and s is a parameter related to energy release rate. Based on a blast model fit with b fixed at 0.6 (s = 1, n = 3),¹ the shock wave should reach the ECQCL beam position in ∼65 μs. Clearly, the blast and drag model fits are only approximations given the limited number of measured data points, but they do provide a range of expected time and length scales for the early fireball dynamics.

We interpret the initial dynamics of absorption and emission signals shown in Fig. 5 as driven by the fireball expansion. After detonation, the fireball immediately begins emitting broadband radiation including light in the MWIR region. The field of view of the detector for emission signals is limited by the window apertures in the chamber, but emission is detected immediately after breakout due to reflections from the chamber walls. As the fireball expands to enter the detector's field-of-view at times of ∼50–100 μs, the MWIR emission signal increases dramatically. For some experimental runs, for example, Comp B in Fig. 5(b) and PETN in Fig. 5(c), the rise in MWIR emission appears delayed. It is possible that for these runs, the fireball radius stagnated before fully entering the detector field-of-view. Given the results in Fig. 23, the expected radius for the fireball stagnation is very near the distance from charge to the measurement beam path and small variations in the initial expansion could lead to variations in arrival time.

Detection of a change in absorption is in all cases delayed relative to detection of emission. In most cases, the initial signal detected in absorption is a transient spike with a short duration of <10 μs. This observation is consistent with a shock wave propagating through the ECQCL beam path and causing a transient deflection of the beam to reduce the intensity reaching the detector. A shock wave with velocity of 4 km/s would travel a distance corresponding to the 0.4 cm beam diameter in 1 μs, but would continue to deflect the beam for longer times as the spherical shock wave intersects the line of the beam path. The arrival time of the initial spike in the absorbance signal ranged from 70 to 140 μs for Comp B, PETN, and PBXN-5, which is the correct order of magnitude for the predicted shock wave arrival time at the beam path. The longer delay observed for NM-AP may be due to the longer distance between charge and the beam path, or from a weaker shock wave due to energy lost in fragmenting the plastic holder surrounding the explosive.

After the initial shock wave passes the beam path, some experimental runs showed a region of almost complete transparency aside from additional transient spikes. We interpret this time period to correspond to after the initial shock wave has passed the beam path, but before the fireball has arrived. It is interesting to note that the spikes observed in absorbance attributed to shock wave propagation to do not appear to correspond with increased signals in emission. Furthermore, the MWIR emission signals evolve slower than the absorption. Since the absorption and emission signals were recorded using the same detector at exactly the same sampling rate, we conclude that the propagation of the shock wave does not produce substantial MWIR emission due to shock-heating of the air. Instead, it appears that the emission evolves over longer time scales and the magnitude of emission is correlated with times of increased absorption. This observation is consistent with the MWIR emission arising primarily from heated, solid particulate matter in the fireball. Times at which a large emitting particle density is present in the beam path thus correspond to a decreased ECQCL transmission from scattering or absorption.

The fireball opacity immediately after detonation is difficult to evaluate with the current setup, due to uncertainty in the fireball intersection with the beam path. It is apparent from Fig. 6 that absorption from CO and CO₂ may appear by 1 ms after detonation, and light resonant with these transitions will be highly attenuated due to the high molecular number density. Figure 8 showed that away from molecular absorption lines, the mean absorbance over the first 10 ms was 2. Although we do not know the exact fireball diameter at this time, we can assume it is less than 1 m as the measurements show the gases have not filled the chamber until t > 1 s. Thus, we place a lower limit on fireball opacity of 2 m⁻¹. For the small 14 g charges used in these experiments, the opacity in the MWIR was low enough to permit transmission measurements throughout the fireball evolution with good SNR except for the NM-AP.

Absorption spectra measurements indicate a rapid decrease in broadband absorbance and temperature within the first 100 ms after detonation. The fireball cools to a temperature <500 K within 100 ms, after which the rate of cooling slows dramatically. Likewise, the broadband absorbance decays quickly in the first 100 ms, followed by a much slower decay over a few seconds. The observed turbulence effects on the beam transmission show changes over similar time scales. Between ∼0.1 and 10 s, mixing of the detonation products with the ambient gases in the chamber continues to decrease the gas and particle number density along the ECQCL beam path, and the temperature decreases via convection. The measurements show that although the rates of change of gas and particle density have slowed, complete equilibrium has not been reached even 100 s after detonation.

The MWIR emission signal was detected up to times ∼10–20 ms after detonation. Based on simulations of blackbody radiance integrated over the MWIR band of 3–5 μm, it is estimated that the detector is not sensitive to emission from sources with T < 500 K. Gas temperature measurements showed that the temperature dropped below 500 K within 100 ms after detonation, which is consistent with the lack of emission signal at these times. Overall, the observed behavior of the emission signal is consistent with a cloud of particles with initial temperatures >500 K, which cools as the cloud expands and mixes with the ambient gas in the chamber.

The detonation gases CO, CO₂, H₂O, and N₂O were readily detected using the absorption features within the ECQCL scan range used, and the band profiles provide a measurement of temperature via the relative peak areas. In a traditional tunable laser absorption spectroscopy measurement using a narrow spectral range, it would be difficult or impossible to measure lines from all these species in addition to the large broadband absorbance baseline. Furthermore, measurement of temperature requires multiple lines of each species within the laser tuning window. The broadband spectroscopy approach of the ECQCL measurement thus offers a significant advantage in being able to measure all the species simultaneously, in addition to temperature and straightforward determination of the broadband absorbance baseline. In this sense, the ECQCL spectroscopy approach is similar to FTIR-based spectroscopy. However, the ECQCL provides a combination of high spectral resolution, high spectrum acquisition speed, and high spectral radiance, which cannot be achieved with FTIR spectroscopy.

The spectral scan rate of the ECQCL in the current experiments was set to 100 Hz over a scan range of 250 cm⁻¹. These scan parameters provided a compromise between scan resolution, scan range, and scan speed. The current setup was ultimately limited by the speed of the digitizing system (2 MHz) placing an upper limit on the QCL current modulation rate of 500 kHz, which sets the time per spectral point to 2 μs. For the 100 Hz scan, this sampling rate provided 2000 points/scan with an average spacing of 0.125 cm⁻¹, which allowed resolution of the gas absorption lines but was slightly undersampled. Faster scan rates up to 1 kHz are possible with the current ECQCL hardware but require either reducing the scan range or sacrificing scan resolution. Future work will explore a larger parameter space available for scanning to enable high-resolution spectroscopy over reduced scan windows, or faster broadband spectroscopy with lower spectral resolution. Nevertheless, the current spectral acquisition rate of 100 Hz over 250 cm⁻¹ with <0.3 cm⁻¹ spectral resolution allowed measurement of gas concentrations, broadband absorbance, and temperatures in explosions, which was previously unexplored in the infrared spectral region.

Fitting of the absorbance spectra over the large wavenumber range presents challenges, especially considering the variable temperature. Not only are there thousands of spectral lines to simulate at each iteration of the fitting algorithm but the relative peak areas change with temperature in a nonlinear manner. In particular, the variable temperature makes it impractical to use a faster and more robust linear least squares fitting with precalculated reference spectra. The WNLS algorithm used here contains approximations that introduced errors in the parameters determined from the fits and could be improved, although at the expense of longer computation times.

The high sensitivity, precision, and speed of the ECQCL-based measurement allow comparisons of measured parameters over time and between measurement runs, which provides value even in the presence of systematic errors. The in situ measurement also eliminates potential errors due to sampling and postanalysis. The ultimate limitation in accuracy may arise from the path-integrated measurement configuration, although the ability to measure conditions inside the fireball and at low temperatures provides valuable information not obtainable using emission-based techniques. Absorption-based measurements are typically more accurate than emission-based measurements, which are difficult to calibrate radiometrically and have poor accuracy under conditions of high optical depth often experienced in high-explosive detonations. In cases of high optical density in explosive fireballs, emission may originate from surface layers and not be characteristic of conditions inside the fireball. High optical density can lead to the distortion of spectral peak heights or areas through self-absorption, leading to errors in parameters determined from spectral fits. The absorbance values >2 observed at early times of fireball evolution, especially near molecular absorption lines, would likely not meet optically thin conditions required for quantitative emission-based measurements. Emission spectroscopy also measures upper state populations of transitions, which may not be in thermal equilibrium with the ground state if formed as products of exothermic reactions as may happen in flame chemiluminescence.

In this manuscript, we reported measurements of the MWIR absorption and emission properties of high-explosive detonations using swept-ECQCL spectroscopy. Fast temporal sampling at 2 μs provided information on MWIR absorption and emission properties during the early stages of detonation, including shock wave propagation, fireball expansion, and turbulence. Spectral measurements at a 100 Hz rate were used to measure gas concentrations of important combustion species CO, CO₂, H₂O, and N₂O. In addition, a broadband attenuation from particulates was measured throughout fireball evolution with a spectrally flat (gray-body) absorption coefficient.

The measurements reported here provide an initial investigation into infrared absorption spectroscopy of high explosives and open many possible future studies. Repeated measurements should provide additional information on statistical variations and uncertainties in parameters for different explosive types. Variation of the measurement configuration, especially the beam distance from the initial charge, should provide additional information on the initial fireball expansion dynamics. Correlation of absorption and emission signals in the IR to measurements in UV/VIS/NIR spectral regions will be of great value.

There are also a wide range of parameters that can be varied regarding the ECQCL scans, to study different aspects of the high-explosive detonation. Higher spectral resolution at slower scan rates would improve measurements of gases at late times after detonation. Faster scan rates will improve measurements at early times, although possibly at the expense of spectral resolution or scan range. Extending the studies to other wavelength ranges, especially the LWIR, will provide information on other gaseous species and the absorption properties of particulates.

This work was supported by the National Nuclear Security Administration, Defense Nuclear Nonproliferation R&D Office. The Pacific Northwest National Laboratory is operated for the U.S. Department of Energy (DOE) by the Battelle Memorial Institute under Contract No. DE-AC05-76RL01830.